摘要
利用多种机器学习算法构建四大家鱼产卵量预测模型,为预测四大家鱼产卵量及评价生态调度效果提供依据。以长江中游宜昌到枝江段为研究区域,基于2012年四大家鱼生态调度期间的监测数据,利用River 2D和CASiMiR-Fish构建水力学模型和栖息地适宜度评价模型,再基于3种栖息地评价方法(几何平均法、算术平均法和乘积法)计算栖息地加权可利用面积,结合监测的涨水量和产卵量数据,利用多种机器学习算法构建四大家鱼产卵量预测模型,通过精度评价选取最佳的栖息地评价方法和机器学习算法组合用于产卵量的预测。结果显示,水力学模型计算的水深和流速能较好地拟合实测数据,当河道流量在10000~15000 m^(3)/s时产卵栖息地面积最大,利用几何平均方法开展栖息地适宜度评价,并利用广义可加模型GAM可以得到最佳效果,预测结果的平均相对误差为28.6%。
Water conservancy projects have affected water temperature,river hydrology,and hydraulic conditions in the middle reaches of the Yangtze River,consequently affecting the natural reproduction of the four major Chinese carps.Reservoir ecological management is an important measure to more closely match natural hydrological conditions and promote spawning of the four major Chinese carps.In this study,the Yichang⁃Zhijiang river section of the middle Yangtze River was selected for study,and predictive models for the spawning volume of the four major Chinese carps were developed under ecological dispatch scenarios.Our aim was to provide a basis for predicting the spawning volume of these carps and evaluate the effectiveness of ecological water regulation.The study was based on monitoring data collected during the ecological water regulation period for the four major Chinese carps in 2012.The hydraulic model and a habitat suitability evaluation model were developed using River 2D and CASiMiR-Fish.The weighted useable habitat area(AWU)was calculated based on three habitat assessment methods(geometric mean,arithmetic mean,and product methods).Then,combining the data on water level rise and spawning volume,we employed various machine learning algorithms to build predictive models for the spawning volume of the four carp species.After accuracy evaluation,we selected the optimal combination of habitat assessment model and machine learning algorithm for predicting spawning volume.The habitat suitability assessment conducted using the geometric mean method,in conjunction with the Generalized Additive Model(GAM),yielded the best results,with an average relative error of 28.6%in the predictions.The water depth and flow velocity calculated by the hydraulic model well fit the measured data,and when the river flow was in the range of 10000⁃15000 m^(3)/s,the spawning habitat area was the largest.We recommend using the geometric average for evaluating habitat suitability,in conjunction with the Generalized Additive Model(GAM).This study provides methodological support for predicting spawning production of the four major Chines carps under ecological management scenarios.
作者
李杨
林东升
张杨
唐金勇
金瑶
杨志
任玉峰
张鹏
常剑波
LI Yang;LIN Dongsheng;ZHANG Yang;TANG Jinyong;JIN Yao;YANG Zhi;REN Yufeng;ZHANG Peng;CHANG Jianbo(State Key Laboratory of Water Resources Engineering and Management,Wuhan University,Wuhan 430072,P.R.China;Chongqing Jialingjiang Lize Avionics Development Co.,Ltd.,Hechuan 401500,P.R.China;Key Laboratory of Ecological Impacts of Hydraulic-Projects and Restoration of Aquatic Ecosystem of Ministry of Water Resources,Institute of Hydroecology,Ministry of Water Resources and Chinese Academy of Sciences,Wuhan 430079,P.R.China;China Yangtze Power Co.,Ltd,Yichang 443000,P.R.China)
出处
《水生态学杂志》
北大核心
2025年第6期71-80,共10页
Journal of Hydroecology
基金
国家自然科学基金项目(U2240213)
湖北省技术创新重点研发计划项目(2023BCB110)
重庆嘉陵江利泽航电开发有限公司科研项目(LZ-DL-2024-021)
中国长江电力股份有限公司科研项目(242102024)。
关键词
四大家鱼
栖息地模型
机器学习
产卵量预测
four major Chinese carps
habitat model
machine learning
spawning prediction
